Dec., 1941
MOLECULAR STRUCTURE O F ALUMINUM DIMETHYL HALIDES
3287
[CONTRIBUTION FROM THE CHEMISTRY LABORATORY OF T F ~ EUNIV~RSITY OF MICHIGANAND THE GEORGE HERBERT JONES LABORATORY OF THE UNIVERSITY OF CHICAGO]
The Molecular Structures of the Dimers of Aluminum Dimethyl Chloride, Aluminum Dimethyl Bromide, and Aluminum Trimethyl and of Hexamethyldisilane AND N. R. DAVIDSON' BY L. 0. BROCKWAY
The problem of the nature of the binding in aluminum trimethyl in the vapor state has aroused the interest of structural chemists since vapor density measurements indicated the existence of double molecules. The most recent vapor density determinations2 show that the trimethyl is completely dimerized a t temperatures below 70" and pressures close to the saturation values, while the triethyl compound is 12% associated a t 150.6'. The present electron diffraction investigation was undertaken in order to compare the structures of the double molecules of aluminum dimethyl chloride and bromide with that of aluminum trimethyl and of the latter with the molecule hexamethyldisilane, since it appeared possible that these two substances would have similar geometrical structures. For the latter compound a new preparation and vapor tension data are reported. A previous electron diffraction study of aluminum trimethyl was made in another laboratory on a sample prepared by one of the present authors* but a reinvestigation with the aid of photographs extending to larger scattering angles is reported here. Experimental Aluminum trimethyl was prepared in the manner described by Laubengayer and Gilliam.2 The preparation of aluminum dimethyl chloride and bromide and the vapor density determinations which show that these substances in the vapor state near room temperature exist as double molecules are described in detail elsewhere.' Hexamethyldisilane has been prepared previously from hexachlorodisilane and the methyl Grignard reagent.b.6 An attempt was made to adapt the method of Friedel and Ladenburg,' who prepared hexaethyldisilane by gently warming a mixture of zinc diethyl and hexaiododisilane. We observed that zinc dimethyl does not react with hexachlorodisilane even a t 130"; in the presence of ethyl ether, however, the reaction proceeds smoothly at 80". (1) Present address: Dept. of Chemistry, University of Mirhigan, Ann Arbor, Michigan. (2) Laubengayer and Gilliam, THIS JOURNAL, 63, 477 (1941). (3) Davidson, Hugill, Skinner and Sutton, Tyans. F W Q ~ Q YSoc.. 36, 1212 (1940). forthcoming publica(4) Brown and Davidson, THISJOURNAL, tion. ( 5 ) For early preparations, see Krause and Von Grosse, "Die Chemie der metall-organischen Verbindungen," Gebriider Borntraeger, Berlin, 1937,p. 268. (6) Schumb and Saffer, THISJOURNAL, 61, 363 (1939). (7) Friedel and Ladenburg, Ann., 303, 251 (1880).
To 3.71 g. of freshly distilled SizClswas added a 31% excess (1216 cc. S. T. P.)of zinc dimethyl vapor and 222 cc. (S. T. P.)of diethyl ether.8 The reaction tube containing the mixture was sealed off and heated for several hours a t 85". A grayish-black, finely divided powder, having the appearance of zinc dust, and two liquid layers formed. After reconnection to the vacuum line, the mixture was dis. tilled off, leaving a large volume of white solid, which probably consisted chiefly of the ether complex of zinc chloride, ( (C2H')20)2.ZnC12. The hexamethyldisilane was separated by repeated distillations from a 20' bath through a trap a t -50' into a receiver at -80". The final sample was tensimetrically homogeneous as shown by careful vapor tension measurements on various fractions. The yield was 1.49 g. or 74% of thestheoretical. The vapor tension data are given in Table I.
TABLE I VAPOR TENSION AND THERMAL DATAON HEXAMETHYLDISILANE 1 , '2.
20.3 26.4 32.9 38.9 42.2 44.0 47.0 54.5 61.2 P , m m . 22.9 31.2 42.4 56.9 66.2 70.4 80.6 109.1 141
+
log,, p = -1920/T 7.907; mean dev. of calcd. fJ from expt. fJ = 0.8%; b. p. (extrapolated) = 109'; AH (vaporization) = 8.8 kcal.; m. p. (observed) = 14.014.4'; Trouton constant = 23.1. The boiling point has been reported' as 112.5". The molecular weight observed for a 89.5-mg. sample occupying 13.4 CC. (S. T. P,)was 149 (hfcslcd. = 146.6). The electron diffraction photographs were made in an improved cameraQwith a plate distance of 10.34 cm. and an electron wave length of 0.0592 A. The four compounds were photographed at temperatures of 20 to 30", with vapor pressures of 5 to 15 mm. for the three aluminum compounds and of about 25 mm. for silicon hexamethyl. At these temperatures and pressures the three aluminum compounds exist in the vapor state entirely in the form of double molecules, i. e., with two A1 atoms per molecule.
In the following presentation of results the observed SO values (equal to 47r (sin 0/2)/X) for each measured maximum and minimum are tabulated together with the coefficients, ck, derived from the visually estimated intensities, Ik. The radial distribution functions, D ( r ) = c c k (sin skY)/skr, were calculated with ck = s2k exp(-mk2)Ik and with exp (-usz) = 0.1 for the largest ring. The resulting curves shown in Fig. 1 have a series of maxima which represent various interatomic distances in the respective molecules. In general the ( 8 ) The vacuum apparatus and techniques required in this preparation are described in ref. 4. (9) Eyster, Gillette and Brockway, THIS JOURNAL, 63, 3236 (1940)
L. 0. BROCKWAYAND N. R. DAVIDSON
3288 -
1
I
1
I
I
I
n A
I
H3)4Brz
Vol. 63
Aluminum Dimethyl Chloride.-Nine maxima were measured on the photographs of aluminum dimethyl chloride with so values ranging out to 19.7. The experimental data in Table E 1 give a radial distribution curve showing strong peaks a t 2.32 and 3.37 8. A consideration of the number of terms and their relative coefficients in the scattering function suggests that these two distances probably belong to Al-Cl and to a mean value of the C-CI and C1-C1 terms. TABLEI1 ALUMINUM DIMETHYL CHLORIDE (DIMBR) Max. Min.
1 2 2 3 3 4 4
5 5 6 6
7 7 8 8 9 9
So
Ck
2.29 3.07 4.06 5.12 6.03 7.05 7.94 8.76 9.67 10.50 11.53 12.68 14.04 15.47 16.87 18.41 19.71
10 - 4 30 -45 +44 -41 +35 -10 16 -34 36 -46 24 -17 11 -4 4
Average Averagedeviation
I 1.5
I I 1 I 3.0 3.5 4.0 4.5 r, A. Fig. 1.-Radial distribution functions. The vertical lines show the relative scattering power associated with the interatomic distances in the final models. I
I
2.0
2.5
values obtained from the two most prominent peaks are reliable to one or two per cent. The tables also show the ratios of stheor/sobs for the measured rings where the values of Stheor are taken from the indicated intensity curves calculated with the atomic scattering factors replaced by the atomic numbers. The average values of these ratios for the better models of each compound are multiplied by the assumed interatomic distances to give the experimental values reported. The agreement between the photographs and the simplified type of intensity curve used here has been shown by tests on molecules of known structure to be good for rings with s values larger than about 4.0. For that reason rings with lower s values are excluded in calculating the average Stheor/Sobs values.
Sr/so
0.958 0.956 .988 ,980 1.011 1.005 0.997 0.999 1.002 .999 0.983 1.000 0.990 ,972 1.022 1.014 0.989 .969 ,992 .015 2.301 3.383
Al-Cl,A. c-Cl,A. Al-Cl 2.31 * 0.03 A. C-Cl 3.43 * 0.05 A. LClAlCl 89 *.'4
-- -
S6/SO
%/SI
si/%
SP/SO
0.965 0.968 0.968 ,982 .973 .980 1.003 .998 1.001 0.994 ,984 1.000 1.015 1.020 1.031
.998 0,998 1.006 .997 0.985 1.000 ,997 .989 0.986 ,987 .984 ,964 ,992 1.004 1.005 1.018 1.014 1.009 1.013 1.002 0.984 0.984 1.010 1,009 0.993 0.996 0.994 ,014 .011 ,014 2.304 2.311 2.306 3.416 3.426 3.459 A1-C = 1.85-2.00A. LCAlC 120-135'
-
0.985 ,985 ,982 1.003 1.025 1.002 0.972 0.995 ,016 2.308 3.423
The principal models considered were based on the bridged structure having a four-membered ring of two AI and two C1 atoms as reported for A12Cle, AlzBrs and A ~ z I vapors.1° s Two methyl groups (with C-H = 1.09 8.and L HCH = 109.5") were attached to each aluminum in a plane normal to the four-membered ring. Preliminary calculations showed that of the terms involving scattering by hydrogen atoms only those associated with the shortest C-H and Al-Hdistances had any detectable effect on the intensity curves, and only those hydrogen terms were included in calculating the curves. The models also were chosen to have strong scattering terms associated with the distances 2.32 and 3.37 A. as required by the radial distribution function. Within this limitation a range of parameter values was tested as shown in Table 111. The extreme values of the bridge angle a t the aluminum atom, i. e., 100 and SO", both give unsatisfactory agreement with the photographs (IO) Palmer and Elliot, TSIS JOURNAL, 60,1852 (1938).
MOLECULAR STRUCTURE OF ALUMINUM DIMETHYL HALIDES
Dec., 1941
3289
TABLE 111 BRIDGEDMODELSOF
ALUMINUM
DUXETHYL CHLORIDE
(DIMER) Model
1 2 3 4 5 6
7 8 9 10
L ClAlCl
L CAlC
looo 93 93 90 90 90 90 90 86 80
120O 128
In all models, AI-Cl = 2.32
128
148 136 125 120
117.2 127.6 139.4
A. and C-H
Al-c,
A.
1.87 1.84 1.90 2.10 2.00 1.90 1.90 1.80 1.90 1.85 = 1.09
100.- I
93'-3
90.- 4
90'-5
A.
since for the 100" model the fourth and fifth maxima (near s = 8 and 9) are shown to be about equal, while in the 80" model the fifth maximum is missing. The best agreement is obtained for the 90" bridge angles with the C-Al-C angle lying between 120 and 135" (Models 5,6, and 7, Fig. 2). Models 3 and 9 with bridge angles of 93 and 86" are almost as good, although Model 9 gives poorer agreement than 3 a t the seventh maximum near s = 14. The quantitative comparisons of the theoretical and observed positions of the maxima and minima are given in Table I1 for Models 3, 5, 6, 7, and 9. At the foot of the table are shown the average values of the ratios of theoretical and observed s values and the values of the two strongest scattering terms obtained by multiplying the observed ratios by the distances assumed in the various models. The resulting parameter values are: AI-C1 = 2.31 * 0.03 8.,C-C1 (non-bonded) = 3.43 * 0.05 k., LClCCl = 89 * 4" with the AI-C bond length probably lying between 1.85 and 2.00 A. and the C-AI-C angle between 120 and 135". One intensity curve was calculated for an ethane-like model in which each aluminum atom is bonded to two methyl groups, one chlorine atom and one aluminum atom, and the qualitative and quantitative agreement with the observed pattern were both poor. A systematic test of models of this type was not made for the chlorine compound, but was made for the bromine compound. We believe that the unsatisfactory character of the ethane-like models in the latter case makes such models improbable in the chlorine compound. Aluminum Dimethyl Bromide.-Photographs of aluminum dimethyl bromide showed eight rings of which the fourth and fifth form a partially resolved doublet with the fifth much weaker than the fourth. The radial distribution
90'-7
900-8
86'-9
80'-10
i
I
1
I
I
5
10
15
20
S. Fig. 2.-Theoretical intensity curves for models of AI?(CH&C12 (see Table 111). The vertical lines show the observed SO values compared with the maxima and minima of the best molecular models.
curve based on the data of Table IV shows two strong maxima a t 2.42 and 3.52 8. Models based on the bridged structure have the most important contributions to the diffraction pattern associated with the AI-Br bond distances and the non-bonded C-Br and Br-Br distances. The first of these is fixed a t 2.42 k. by the radial distribution function and the average of the other two was made equal to 3.52 k. in each of the models listed in Table V, except Model 9, for which the corresponding values were taken a t 2.35 and 3.55 8. The models with the LBrAlBr smaller than 90" lead to curves showing only one maximum near s = 8 (e. g., Curve 8, Fig. 3) where two are observed on the photographs. Curves based on models with a bridge angle larger than 90" show an extra ring a t s = 14 (Curve 2, Fig. 3). A value of 90" for the bridge angle leads to a set of qualitatively correct curves with C-Al-C angles ranging from 115 to 130". Outside of this range the appearance of the doublet near s = 8 no longer agrees with the observed pattern.
L. 0. BROCKWAY AND N. R. DAVIDSON
3290
93'-2
90'- 3
90'-4
90'-6
90'-7
V 5
10
15
S. Fig. 3.-Theoretical intensity curves for models of Ale(CH&BrP (see Table V).
An ethane-like model formed by joining the aluminum atoms of two AI(CH&Br groups was considered as a possibility because a model of this type is reported for Alz(CH3)6. The range of values of the molecular parameters is limited by the following considerations based chiefly on the radial distribution curve. The 2.42 8. distance is again taken to represent the AI-Br bond length and the peak a t 3.52 A. is due to Br-Br, C-Br and the nonbonded A1-Br distances. The Al-AI distance ought to lie between 2.0 and 2.4 8. with C-AI-Br angles somewhat larger than 110" as in the case of the corresponding parameters in aluminum trimethyl. The Br-Br distance then lies between a minimum value in the cis configuration of 3.0 * 0.2 A. and a maximum value in the trans configuration of 5.6 * 0.2 8. The radial distribution curve allows an intermediate value of about 3.5 A., which fixes the molecule with one end rotated about 40" from the cis position. In this model
Vol. 63
two AI-Br distances occur, 2.42 and 3.54 A., and also two C-Br distances, 3.50 and 5.20 8. The first three values agree well with the positions of the two most prominent peaks in the radial distribution curve, but for this model the second of these peaks should be more than three times as strong as the first according to the relative scattering power associated with each. The intensity curve based on the model just described shows a series of regular maxima and minima whose positions agree fairly well with those of the measured rings on the photographs, but the distinctive qualitative features of the observed pattern do not appear at all in the curve. In particular, the double maximum near s = 8 (as in Curve 4, Fig. 3) appears as two completely resolved maxima with the second stronger than the first. A further test was made by using another ethane-like model in which the distances a t 2.42 A. and near 3.50 8. were retained in order to preserve the agreement with the radial distribution function, but the bond angles all were set a t 109.5", the Br atoms in the cis-position, the AI-C bond distances and the AI-A1 bond a t 1.89 8. The resulting curve is indistinguishable from the foregoing one. This will be true for any other ethane-like model which agrees with the radial distribution function because the major part of the scattering power of the molecule is associated with the two important maxima in this function, and any variations among the remaining terms have no appreciable effect on the calculated intensity curves. A bridged model, on the other hand, has been found which not only gives agreement with positions of the peaks in the radial distribution curve but also better represents their relative importance as well as showing good quantitative and qualitative agreement with the calculated intensity curves. We believe, therefore, that the ethane-like models must be rejected in favor of a bridged model with 90" angles. The quantitative comparison of Table IV leads to the final results: A1-Br = 2.42 * 0.03 8.; C-Br = 3.59 * 0.05 A.; LBr-AI-Br = 90 * 3"; LC-AI-C = 115-130" and AI-C = 1.902.05 8. Aluminum Trimethyl.-The two innermost maxima were very strong, the second being a little broader than the first and showing a slight asymmetry on the outer edge. The third and fourth were only partially resolved with the latter somewhat weaker. The fifth ring was sharp and
MOLECULAR STRUCTURE OF ALUMINUM DIMETHYL HALIDES
Dec., 1941
3291
well-defined, while the sixth was rather broad. A seventh ring was observed, but measurements S6/SO Max. Min. so Ck SJSO sdso on it were not reproducible. The observed SOval2.19 9 (1.106) (1.110) (1.106) 1 ues are shown in Table VI. The first nine values 2 3.07 - 9 (0.961) (0.977) (0.961) correspond to the maxima and minima reported .972 2 3.95 24 .977 .980 by S ~ t t o n whose ,~ values average about 0.5y0 .981 .979 .986 3 4.92 -38 smaller than ours. Our radial distribution func,995 ,988 1.000 5.83 37 3 .990 0.993 4 6.85 -45 .987 tion based on twelve maxima and minima shows 1.007 1.013 4 7.74 33 1.019 two strong peaks a t 2.01 and 3.30 K., with a 5 8.43 - 4 weaker intermediate one at 2.58 A. (Fig. 1). 1.014 0.999 5 9.13 11 0.997 The first tests were made on models of the 1.007 1.006 1.010 10.00 -30 6 ethane type with trigonally symmetrical alumi1.008 1.013 1.008 30 6 10.97 0.995 1.003 0.999 12.16 -14 7 num trimethyl groups joined by the aluminum 1.019 7 13.17 13 1.025 1.015 atoms. Changes in the relative orientation of 15.36 3 8 the two trimethyl groups about the axis of the 0.999 1.002 0.997 Average molecule had no effect on the intensity curves. 0.013 0.011 0.011 Average deviation For that reason in all but three models a single 2,424 2.414 2.418 AI-Br, 8. orientation was used, i. e., with the carbon atoms C-Br 3.597 3.576 3.589 a t one end of the molecule lying in planes through AI-Br = 2.42 * 0.03 A. A1-C = 1.90-2.05A. LC-A1-C = 115-130' the axis which bisected the C-A-C angles a t the C--Br = 3.59 += 0.05 A. LBr-Al-Br = 90 * 3' other end. The long C-H distances were usually neglected since their contributions had no detectTABLE V able effect on the intensity curves. The models BRIDGEDMODELSOF ALUMINLTM DIMETHYLBROMIDE calculated are described in Table VI1 in terms of Model L Br-Al-Br L c-AI-c A1-C, A. two parameters, the C-AI-Al bond angle and the 1 95O 120' 1.83 ratio of the AI-A1 to A1-C bond lengths, following 2 93.4 120 1.90 the scheme used by Sutton. The absolute scale 3 90 109 1.85 4 90 115 1.90 was fixed in these models by assuming an AI-C 5 90 120 1.90 bond distance of 2.00 A. as suggested by the 6 90 126 2.00 radial distribution curve. Intensity curves for 7 90 136 2.10 Models 1 to 8 are shown in Fig. 4. All of these 8 86.5 113 1.95 curves except 5 show the partially resolved third 9 87 115 1.95 and fourth maxima although only in 1, 2 , 6 and 7 Models 1-8, AI-Br = 2.42, C-H = 1.09 A. Model 9, AI-Br = 2.35, C-H = 1.09 A. is the fourth ring weaker than the third as obTABLE IV ALUMINUMDIMETHYL BROMIDE (DIMER)
TABLE VI ALUMINUM TRIMETHYL (DIMER) Max.
Min.
2
3 4
5 6
so
2.20 2.89 3.73 5.33 6.31 7.14 7.88 8.98 10.00 11.43 12.72 (17.39)
Average Average deviation All-C A12-Ci AL-CI = 2.01 * 0.04 8. A11-C = 3.24 * 0.06 .&.
Ck
7 - 8 19 -28 29 - 7 27 -44 37 -24 9 3
Si/SQ
S 6 h
SII/SO
SldSO
SlS/SO
(1.082) (1.059) (1.085) 0.976 1.038 1,018 0.977 0.991 1.030 1.020 1.037
(1 ,068) (1.038) (1 .067) 0.957 1.022 1.001 0.983 0.989 1.018 0.983 1,019
(1.105) (1,100) (1.080) 0.970 1.017 1.008 0.977 0.980 1.027 1,010 1.018
(1.077) (1.067) (1.072) 0.947 .995 .983 .983 .978 1.030 0.993 1.038
(1.109) (1.090) (1 .094) 0.964 1.005 1.003 0.996 0.997 1.305 1.008 1.060
1.013 0.020 2.026 3.252
1.002 0.015 2.004 3.296
1.005 0.015 2.010 3.196
1.000 0.019 2.000 3.260
1.015 0.019 2.030 3.258
LC-AI-AI
= 100
* 6".
L. 0.BROCKWAY AND N. R. DAVIDSON
3292
VOl. 63
109'- I
100'-10
109'-2
100'-11
109'-3
100'- 12
100'-13
109'-4
95'- 14 105'- 5 95'-15
105'- 6 95'- 16
105'- 7
U 105'- 8
I
I
I
5
10
15
3.
V I
I
I
5
10
15
S. Fig. 4.-Theoretical intensity curves for 109' and 105' models of Alz(CHJe (see Table VII).
served in the photographs. The 109.5' curves all show an extra ring between the observed fifth and sixth. Curves 6 and 7 are somewhat better in this region, but the double maximum shown here appears on the plates as a single broad ring. From these observations we conclude that the C-A1-Al angle is definitely less than 109.5' and probably less than 105'. I n Fig. 5 some of the curves for 100 and 95' models are shown. Curves 9 and 10 do not show the third and fourth maxima resolved. Curves 11 and 12 give the best reproduction of the cha-racteristic features of the diffraction pattern of all the calculated curves. In curve 13 the fourth maximum is too much stronger than the third. Of the 95' curves number 15 is the best, but the sixth ring here is less like the photographs than is the corresponding ring in curves 11 and 12. From the foregoing comparisons it appears that
Fig. 5.-Theoretical intensity curves for 100' and 95' models of Al*(CH& (see Table VII).
for a given angle better curves are obtained with small values of the Al-Al/Al-C bond length ratio when the angle is large and with larger length ratios when the angle is smaller. It is found that the more satisfactory curves for the various angles all have nearly the same value of the ratio of the non-bonded Al-C distance to the bonded A1-C distance, i. e., Al2-CI/Al1-C1 = 1.61 f 0.02. For the ethane-like models these two terms represent about two-thirds of the total scattering power of the molecule. The 90' model indicated in Table VI1 was chosen to have the same value of TABLE VI1 ETHANE-LIKE MODELSFOR ALUMINUM TRIMETHYL (DIMER) LC-Al-A1
90'
95O
100'
105O
109.5'
5 6
1 2 3 4
A1-Al/ A1-C
0.975 1.025 1.075 1.125 1.175 1.225
1.25
14 15 16 17
9 10 11 12 13
7 8
Dec., 1941
MOLECULAR STRUCTURE OF ALUMINUM DIMETHYL HALIDES
3293
the last mentioned ratio as the most promising particularly in the region of the distinctive third curve in the 100' group, but it fails to reproduce and fourth maxima. The parameters of the sixth the observed third and fourth rings. A considera- model were chosen to reduce as much as possible tion of the qualitative agreement between the the difference between the two terms averaging photographs and the curves based on ethane-like 3.3 8. while maintaining an average Al-C bond models leads to a model having L C-A1-Al = 100' length of 2.00 A. The resulting values were and a ratio of the two Al-C distances equal to 1.61. C-C = 3.18 A. and AI-C = 3.38 A. with Al-C The difficulty of accounting for the existence of bond lengths of 1.90 and 2.10 A.a bridge angle the ethane-like molecule in terms of existing bond of 126' and an external angle of 110'. The scattheories led to the testing of other possible models. tering curve fits the third and fourth maxima quite I n particular, it was decided to test the bridged well but the positions of the sixth minimum and structure found in the Al2(CH3)&2 models. Two maximum were in error by-three and seven per AI and two C atoms were placed in a four-mem- cent., respectively. This model is also unsatisfacbered ring with two additional methyl groups at- tory in bringing non-bonded Al atoms within tached to each AI atom in a plane perpendicular to 1.9 A. of each other. the ring. Three models were calculated using The bridged models giving reasonable agreement parameters suggested by the results on the halogen with the radial distribution curve lead to unsatissubstituted molecules: i. e., using a square ring, factory scattering curves (as well as giving iman external C-A1-C angle of 125 or 120' and the probable values to some of the interatomic separafollowing Al-C bond lengths in the ring and ex- tions). Since any other models would necessarily ternal to the ring: Model A - 1.90, 1.90 A,; give poor quantitative fit with the photographs, B - 1.80, 1.90 A.; C - 1.90, 1.80 A. The cor- there is little hope of finding a satisfactory bridged responding intensity curves show extremely poor model on the basis of the present data and we conagreement with the observed pattern. In A, the clude that the most probable structure is reprefourth ring is missing, but two extra rings appear sented by the ethane-like model described above. near the fifth and sixth; in B the second ring is The final parameter values are determined by doubled, the fifth is much too weak, while the sixth comparisons for the better models as shown in appears as an inner shelf on a strong seventh. Table VI. The first three measured so values fall A second group of bridged models was chosen below 4.0 and are excluded from the final average to give a t least fair agreement with the strong because (as usual for the smallest rings) they do radial distribution peaks at 2.01 and 3.30 A. The not give results in agreement with those from the average of the bonded Al-C distances was set at larger rings. The final values are Al-C (bonded) 2.00 8., while the average of the other two im- = 2.01 * 0.04A.; Al2-C (non-bonded) = 3.24 * portant terms, the separation of non-bonded 0.06 A.; L C-Al-AI = 100 * 5'; A1-Al = 2.20 * Al and C atoms and of C atoms in the ring and 0.15 A. outside the ring, was kept near 3.3 A. The reHexamethyldisi1ane.-Seven maxima and five sulting C-AI-C bridge angle is about 120' so that minima were measured in a range of s values the two AI atoms approach each other within from 4.1 out to 20.6. The third ring is stronger about 2.1 A. This close approach of non-bonded than the second, the fourth is sharp, while the Al atoms is very unreasonable but it is not pos- fifth ring is broad and appears to be a double ring, sible to avoid this difficulty in any bridged model but not well enough resolved to afford measurewith distances required by the radial distribution ments on the components. In Table VI11 the curve. Three of the six models were calculated value shown for the fifth ring is the measured with C-AI-C bridge angles of 120' and external center of the doublet. The radial distribution C-Al-C angles of 125'. The external and internal curve based on the data in Table VI11 shows Al-C bond lengths were equal in the first a t 2.00 A. strong peaks a t 1.87 and 2.41 8. with a broad and were set a t 1.90 and 2.10 A. in the second and peak ranging from 3.25 to 3.50 A. a t 1.80 and 2.20 A.in the third. A fourth model A series of ethane-like models was calculated with a 120' bridge angle was calculated with an for the values of bond angles and bond length external angle of 135' and A1-C bond lengths of ratios shown in Table IX. The scale was fixed 1.90 8 and 2.10 A. None of these five models by assuming S i S i = 1.94 A. in each model. As in gave curves reproducing the observed pattern, the case of the aluminum methyl compound, fairly
3294
L. 0. BROCKWAY AND N. R. DAVIDSON TABLEVI11 HBXAMETHYLDISILANE
Max.
Min.
so
Ck
4.07 5.17 6.18 6.98 7.91 9.56 10.91 12.26 13.94 15.60 16.97 20.59
23 -12 22 -26 45 -56 62 -46 34 -13 12 4
Ss/Sa
(0.943) 2 0.977 2 1.021 3 0.984 3 .978 4 .946 4 1.011 5 0.982 5 .986 6 ,990 6 .974 7 .983 Average .986 Average deviation ,013 Si1-G 1.911 Sit-C1 3.448 Si,-Cl (bonded) = 1.90 f 0.02 A. Si& (non-bonded) = 3.47 f 0.04 LC-Si-Si = 109 f 4'. S i S i = 2.34 * 0.10 A. 1
Vol. 63
TABLE IX ETHANE-LIKE MODELS FOR HBXAMBTHYLDIS~LANE
Sr/So
sdso
L C-Si-Si
(0.939) 0.961 1.010 0.978 .964 .953 1.001 0.972 .976 .981 .966 .976 ,976 .012 1.893 3.465
(0.929) 0.948 1.003 0.974 ,967 ,968 ,993 .965 ,971 ,977 .958 ,976 .973 .011 1.888 3.493
Si-Si/Si-C 1.154 1.185 1.206 1.232 1.258 1.294
105'
107'
108'45'
8
10Q028'
113'
2 1
9
7 6
10 11
3 4 5
almost as good. Model 6 with a bond angle of 108'45' was chosen to have the value of the ratio of the two Si-C distances determined from Model 4 and to show the best possible agreement with the radial distribution function. The intensity curves for Models 4 and 6are almost indistinguishable, and therefore a change of only a degree in the bond angle cannot be detected experimentally. A. Model 1 with a bond angle of 113' and Si2-Cl/ Si& = 1.82 gives a curve quite similar to that of Model 4,but a little less satisfactory in the neighsatisfactory intensity curves were obtained with borhood of the third and fourth maxima. Model bond angles covering an eight degree range if the 11, the best of those having 105' bond angles, is ratio of the two strongest scattering terms is kept also only a little less satisfactory than Model 4. nearly constant (at a value of 1.82 for the silicon The final results for hexamethyldisilane and the compound). estimated uncertainties are as follows : Si-C (bonded) = 1.90 f 0.02 A,; Si-C (non-bonded) = 3.47 f 0.04 8.; LC-Si-Si = 109 * 4'; Si-Si = Sc,K ti,), 2.34 * 0.10 A. The relatively large uncertainty in the length of the Si-Si bond is due to the small proportion of the total scattering represented by 113.. I this term.
Discussion 1ov*-3
109*-4
109'-5
6
105*-11
I
I
I
I
5
10
15
20
S.
Fig. 6.-Theoretical intensity curves for models of Si%(CH& (see Table IX). The observed SO values have been multiplied by 0.976, the scale factor obtained for Model 4 in Table VIII.
Among the 109'28' models whose intensity curves were calculated first the best reproduction of the observed pattern is given by 4 with 3 and 5
The results for the four compounds are collected in Table X together with the corresponding values previously reported for AI&& and A12Bre.l0 The AI-C bond distances and angles in the substances A12(CH&C12and A12(CH&Br2are given as ranges of values because within each range there is no experimental reason for preferring one value over another. These two compounds have the bridged structure involving a ring of two metal and two halogen atoms found in the &XG mole~ules,'~ as (11) The possibility of ethane-like molecules was not considered in the invesdgation by Palmer and Elliot. We calculated three intensity curves based on ethane-like models of AIzCls with distances indicated by the two strong peaks in the radial distribution function. The qualitative agreement with the pattern described by the authors was not very good. For AlzBre, however, we found that an ethane-like model with an AI-Br bond length of 2.28 A,, AI-Al 1.90 A. and Br-AI-AI bond angles of 105' gave excellent qualitative and quantitative agreement with the observed pattern. Photographs extending to larger scattering angles would be required for distinguishing between the bridged and ethane-like structures on the basis of electron diffraction data. The bridge structure seems more likely in view of the present result on Ali(CH:)rBrx and of the nutqber of other halogen compounds known to have such structures.
Dec., 1941
MOLECULAR S T R U C T U ~OF
ALUMINUM bIMETHYL HALIDES
3295
TABLE X
c-x
AI-X
A1z(CHs)&lz AldCHdrBrz AlZCle1l AlzBr611
2.31 2.42 2.21 2.33
f
*
0.03 A. 0.03
3.43 3.59
MI-C
Ah(CHs)s Si2(CHdB
2.01 1.90
* 0.04A. * 0.02
AI-C
* 0.05 A. * 0.05 MrC
3.24 3.47
0.04
well as in other metallic halides such as the dimeric diethyl auric bromide, Au2(GH&Br2,12 and the crystalline polymer of palladous chloride, PdC12.13 The Al-X distances appear to be longer in the methyl substituted compounds, in which also there is no evidence for any deviation from 90' angles in the bridge. Since the bonds are probably partially covalent, the formation of the bridge in the Alz(CH3)4Xz molecules presumably involves the occurrence of four electron-pair bonds on aluminum, in one of which both the electrons have been supplied by a halogen atom. Because of the many examples of this type of polymerization no new problem is raised here. In the case of &(CH3)6 the present results confirm the ethane-like structure reported by Sutton and c o - ~ o r k e r s . ~ Our values for the A1-C bond length and the bond angle are given with smaller uncertainties than theirs (2.05 * 0.10 A.and 105 * 10") because the extra ring observed and measured on our photographs made it possible to eliminate certain models. Our consideration of the possibility of bridged models has made the correctness of the ethane-like structures more probable. It is interesting to note that the C-A1-A1 bond angle is definitely less than 109' (the value observed for C-Si-Si). It seems probable that the monomer Al(CH3)s has the coplanar structure observed in B(CH3)314with 120' angles. When two AI(CH3)3 molecules are united, the C-A1-C angle decreases only to about 115'. The dielectric constants and polarizations of the vapors of Alz(CH&C12 and Alz(CH& have been measured by Wiswall and Smyth.15 The latter compound was about 10% dissociated at the temperatures of the measurements and the authors conclude that the monomer Al(CH& must have a dipole moment of about 1.40. The calculations leading to the moment values in the third column of their Table I (p. 353) are based (12) Buraway, Gibson, Hampson and Powell, J . Chcm. SOC., 1690 (1937). (13) Wells, 2.Krist., 100, 189 (1938). JOURNAL, 69, 2085 (1937). (14) Levy and Brockway. THIS (15) Wiswall and Smyth, J . Chem. Phys., 0, 352 (1941).
A. A.
L X-A1-X
L C-A1-C
89 * 4' 90 * 3" 80 O 87 a
12C-135" 115-130" (118') (115')
MI-MI
* 0.06 A. =t
1.85-2.00 1.90-2.05
2.20 2.34
* 0.15 A. 0.10 A.
f
L CMM
100 f 5" 109 * 4"
on the differences between the observed total polarizations and estimated electronic polarizations. These differences are 21 cc. for A12(CH& Clz and 11 cc. for &(CH3)6, and we believe that they are not sufficient to prove the existence of a permanent dipole moment. In A U Z ( C Z H ~ ) ~ B ~ ~ the excess of total polarization over electron polarization is 36.3 cc.12although an X-ray investigationlZhas shown the molecule to have a symmetrical bridged structure (similar to that of Alz(CH3)4C12 except for the planar coordination around the Au atoms) incompatible with the existence of a permanent moment. In AlzBr6 measured in CSZsolutions at concentrations which contain almost exclusively double molecules the excess of total over electron polarization is about 12 cc.16 The probable explanation of these abnormal atom polarizations has been discussed by Coop and S ~ t t o n , who ' ~ expect to find anomalous polarizations in molecules containing groups with large individual moments opposed to each other and in which the individual groups have relatively low force constants with respect to their bending motions. In view of the number of substances cited by Coop and Sutton for their anomalous behavior and the observations by Wiswall and Smyth of excess polarizations for A12(CH3)4C1z and Alz(CH3)6 a t temperatures where the samples are composed mainly of double molecules there is still considerable doubt of the existence of permanent moments. A test of the structure of the monomer Al(CH3)a by measuring the temperature coefficient of polarization a t high temperatures is perhaps not feasible, but we hope to make an electron diffraction investigation of the monomer. For hexamethyldisilane only the ethane-like structure was tested because it is in accord with all of the observed structures of compounds containing four coordinated silicon and because this model gave quite satisfactory agreement with the electron diffraction data. The silicon-silicon bond length is reported as 2.34 A. with an uncer(16) Nespital, Z.phrsik. Chem., B i 6 , 153 (1932). (17) Coop and Sutton, J . Chcm. Soc., 1269 (1938).
3296
L. 0. BROCKWAY AND N. R. DAVIDSON
Vol. 63
tainty of 0.10 A.because of the relative unimpor- the Si-Si bond. Aluminum forming normal tance of this term in the scattering function. covalent bonds in an octet should be larger than The reported value is probably more reliable than Si. An electron deficiency could scarcely shorten indicated, however, since the same bond in Si&& the bond, although it may be noted that if the was observed to be 2.32 =t 0.03 The Si-C electron deficiency does not occur at all in the bond length observed here (1.90 * 0.02 8.) AI-A1 bond but is concentrated in the A1-C bonds, may be compared with the value 1.93 * 0.03 A,, the Al-Al bond might then be shorter than the reported for Si(CH&.lg The difference may not normal octet bonds because the average number be real, but the investigation of Si(CH3)d is being of electrons per A1 atom would be less than eight. repeated on the present apparatus as a further Structures involving only three occupied orbitals test. The radius sum for a single covalent Si-C (sp2) on aluminum would have bonds shorter bond is 1.94 8.,while Schomaker and Stevenson20 than the octet bonds. At the same time the propose a modification of the use of atomic radii Al-C bonds would be relatively longer because of based on the relative electronegativity of the the deficiency there. These suggestions are not bonded atoms which gives for Si-C a calculated very satisfactory because the loss of energy in value of 1.88 A. The bond angles do not difier the electron-deficient AI-C bonds would probfrom the tetrahedral value by as much as 4O, the ably be greater than the energy gained in the formation of the double molecule. The heat of the experimental uncertainty. The hexamethyl derivatives of both aluminum reaction of dissociation is observed to be about and silicon show the same ethane-like structure, 20 kilocalories per mole. That the bond holding but the molecules are differently proportioned. the two aluminum atoms together should be preThe ratio of the two metal-carbon distances is dominantly ionic or polar is unlikely because of the 1.61 in the A1 compound and 1.83 in the Si com- symmetry of the molecule, but electron deficiency pound. The AI-A1 bond is about 10% longer in the A1-C bonds would increase their polar charthan the Al-C bond whereas the S i S i bond is acter and indeed help to account for the abnormal about 25y0 longer than the Si-C bond. If from atom polarization of the substance. the Si-C bond length we subtract a carbon atom Acknowledgments.-The assistance of the radius of 0.77 A., a Si radius of 1.138.is left. This Horace H. Rackham Foundation and of the is to be compared with 1.17 0.05 A., i. e., half of Midgley Foundation which made possible the the observed Si-Si bond. (The agreement here construction of the diffraction apparatus is would be better if the difference in the relative gratefully acknowledged. We are indebted to electronegativities in the two bonds were allowed Dr. H. C. Brown for suggesting the comparison for by the SchomakerStevenson scheme.) I n of Alz(CH& with Si2(CH3)6 and for help in the A12(CH&, subtracting the carbon radius leaves preparation of the latter substance, and to Miss 1.24 A. for AlZ1whereas half of the observed Charlys-Marion Lucht for special assistance in the A1-A1 bond length is 1.10 * 0.08 A. If the bonds preparation and interpretation of the photographs. in Siz(CHa)e are single electron pair bonds, then One of us (N. R. D.) wishes further to acknowledge in the Al compound the AI-C bond is too long helpful advice and criticism from Professor H. I. relative to AI-AI for both to be electron pair bonds. Schlesinger. The existence of double molecules of AI(CH3)3 Summary requires some modification of our bond theories, but the best manner of describing the binding 1. A new preparation of SiZ(CH& is described, forces in this molecule is by no means clear. The and the vapor pressure has been measured over a deficiency of two electrons below the number re- range of temperature from 20 to 60’. quired for electron-pair bonds throughout the 2. The structures of Al2(CH&Clz, A12(CHa)(molecule suggests that the A1-A1 bond should Brz, &(CH3)6 and Si2(CH3)6 have been determined have less than two electrons on the average, but by means of electron diffraction with the results the AI-A1 bond is apparently 0.1 A.shorter than shown in Table X. (18) Brockway and Beach, THISJOURNAL, 60, 1836 (1938). 3. The two methyl aluminum halides have the (19) Brockway and Jenkins, i b i d . , 68, 2036 (1936). bridged structure of two metal and two halogen (20) Schomaker and Stevenson, ibid., 63, 37 (1941). (21) The tetrahedra! covalent radius for A1 is given by Pading atoms in a ring observed in other similar com(“Nature of the Chemical Bond,” Cornel! University Press, Ithaca, pounds and the ring angles are close to 90’. N. Y., 1939,p. 167) as 1.26 A.
CATALYTIC DECOMPOSITION OF AMMONIA
Dec., 1941
4. Al2(CH& and Siz(CH& have ethane-like structures. The silicon compound has the bond lengths and angles expected for a normal covalent structure. In the aluminum compound the C-A1-C bond angles are 115O or larger. The Al-Al
3297
bond is surprisingly short in comparison with the S i S i bond length. 5. The structural question of what makes two Al(CH3)3 molecules join is still a question. ANNARBOR,MICHIGAN
RECEIVED AUGUST 4, 1941
[CONTRIBUTION OF THE FERTILIZER RESEARCH DIVISION,B U R ~ AOF U PLANT INDUSTRY, U. S. DEPARTMENT OF AGRICULTURE]
The Catalytic Decomposition of Ammonia over Iron Synthetic Ammonia Catalysts BY
KATHARINE
s. LOVE AND P. H. EMMETTI
The catalytic decomposition of ammonia has been studied by many workers. In only a few instances, however, have such studies been concerned with the decomposition over iron synthetic ammonia catalysts. The present work had as its primary object the elucidation of the kinetics of ammonia decomposition over several of the many iron catalysts prepared and studied in the U. S. Department of Agriculture. The results obtained point toward a complexity of the ammonia synthesis and decomposition reactions over iron that heretofore has not been recognized. The present work is believed to represent progress in unravelling some of the many complex aspects of these reactions over iron but much additional painstaking work is needed for a complete understanding of ammonia catalysts.
Synthetic ammonia which had been selected for its purity was dried over fused potassium hydroxide. The hydrogen was electrolytic. The nitrogen was atmospheric (Linde) and was p d e d by passage first through one tube of hot copper and after mixing with the hydrogen stream through a second. The hydrogen-nitrogen mixtures were then passed through tubes containing soda lime, through a trap immersed in liquid air and finally through a phosphorus pentoxide tube. The copper had been reduced at 800" and then oxidized and reduced again at a lower temperature. The temperature of the catalyst was measured by a thermocouple placed in a glass well extending through the body of the catalyst. Most of the results reported here are on the decomposition over three iron synthetic ammonia catalysts: Catalyst 931, doubly promoted with 1.3% A I 2 0 8 and 1.59% KzO; Catalyst 954 singly promoted with 10.2% AlzOs; and pure iron Catalyst 973, containing 0.15% Al*O* as impurity. The decomposition was also measured over iron supported on AlzO,.
Apparatus and Procedure
Experimental Results
The apparatus and procedure were similar t o those described by E. Winter." A flow system a t approximately atmospheric pressure was employed. Mixtures of hydrogen, nitrogen and ammonia were led over the hot catalyst. In a few experiments helium was substituted for nitrogen. The undecomposed ammonia was absorbed in half normal sulfuric acid and the excess acid titrated with standard alkali. The rate of decomposition was estimated as the difference between the ammonia flow a t the entrance and at the exit. Each entering gas was measured by a separate flowmeter. Through a multiple capillary arrangement each flowmeter could accurately cover the range 50 to 500 cc. of gas per minute. All of the capillaries were thermostated at 30'. By thus keeping the temperature constant and correcting for pressure changes, satisfactory precision could be obtained. The manometer measuring the flow of ammonia was made from tubing 12 mm. in diameter. It was long enough t o allow a pressure difference of 100 cm. of nujol and was kept at constant temperature. Mercury was used in the auxiliary manometers for measuring the pressure a t the exit of each flowmeter.
Decomposition on Catalyst 93 1.-The initial period of reduction of each sample of 931 included about twenty-four hours a t 400' and twenty-four to seventy hours a t 430 to 450'. The space velocities during reduction were 15,000, 3600, 2400 and 1000 for samples A, B, C and D, respectively. Immediately before each run on this and other catalysts, the catalyst was reduced a t the temperature of the run for a t least one hour. Between runs the cold catalyst stood in hydrogen at a pressure 1.5 cm. of mercury higher than atmospheric. Unless stated otherwise, the data on the decomposition over all of the catalysts were obtained under conditions of temperature and gas composition which preclude formation of the iron nitrides. An apparent energy of activation of 43,800 *
(1) Present address: Department of Chemical and Gas Engineering, The Johns Hopkins University, Baltimore, Maryland. (la) E. Winter. 2. physik. Chcm., 813, 401 (1931).
2s
(2) Emmett, Hendricks and Brunauer, THIS JOURNAL, 63, 1456 (1930). (3) Brunauer, Jefferson, Emmett and Hendricks. ibid., 68, 1778 (1931).
